Colgate says LLM role-play can generate ~90% accurate survey data, challenging market research
ZeroStateReflex · x · 2026-11-22
Colgate published a paper showing LLMs can predict real purchase intent by role-playing target demographics, yielding survey data with roughly 90% accuracy — a quiet threat to the market research industry.
The key is Semantic Similarity Rating (SSR): asking an AI to rate a product 1-5 yields safe, middle-of-the-road answers, so researchers instead give the model a demographic profile and a product concept, have it write raw, unfiltered thoughts, then use a semantic model to convert those texts into a rating distribution. Observers note profound short-term implications and disturbing long-term ones.
More from AGI Musings
- Erik Hoel defends 'qualiaphobia': the moral case for engineering non-conscious AI — erikphoel · 2026-11-22
- Harvard lab uses Biomni AI agents to decode the hidden "grammar" of DNA — anshulkundaje · 2026-11-22
- teortaxesTex cites Yudkowsky's Riemann passage to contest the 'RH-capable AI can kill you' threat model — teortaxesTex · 2026-11-22
- AlphaChip and three TPU generations: a mine that digs itself is a throne — krishnan · 2026-11-22
- Zero QA job searches in 2026 on recruiter tool as AI reshapes testing roles — yenkel · 2026-11-22
- Researcher tells Claude to stop its privacy lectures: "moral bumper bowling" erodes judgment — mattbeane · 2026-11-22